Abstract: In order to monitor foggy weather at low cost and high temporal and spatial resolution,a deep learning-based foggy weather monitoring via wireless communication method was proposed in this paper.Since different concentrations of foggy weather in the channel leave different features in the signal,this paper collects radio signals under four different concentrations of foggy weather to establish the foggy weather monitoring dataset.By introducing an attention mechanism in the conventional ResNet50 network and performing feature fusion,an improved A-ResNet50 model is obtained.The A-ResNet50 network is used to extract the features of different concentrations of foggy weather left in the received signals,and to classify and identify four types of different concentrations of foggy weather for the purpose of monitoring foggy weather.The proposed method was validated on the dataset established in this paper,and compared with other traditional classification algorithms,the network model proposed in this paper has the best performance.The final recognition accuracy reached 86.18 %, and the result proved the feasibility and effectiveness of the method.

Key words: Radio communication, Fog, Meteorological monitoring, Deep learning, ResNet50